Tea-YOLOv8s: A Tea Bud Detection Model Based on Deep Learning and Computer Vision.

Area of Science:

• Agricultural Engineering
• Computer Vision
• Robotics

Background:

• Mechanized selective harvesting of tea requires precise tea bud detection.
• Complex backgrounds in tea fields challenge existing detection models, reducing precision.

Purpose of the Study:

• To develop a novel deep learning model, Tea-YOLOv8s, for accurate tea bud detection.
• To improve the precision and efficiency of tea bud identification in challenging visual conditions.

Main Methods:

• Utilized data augmentation to enhance image quality and quantity.
• Integrated deformable convolutions, attention mechanisms, and improved spatial pyramid pooling into the YOLOv8s architecture.
• Compared the proposed Tea-YOLOv8s model against other mainstream YOLO models.

Main Results:

• The Tea-YOLOv8s model achieved a mean average precision of 88.27%.
• The model demonstrated an inference time of 37.1 ms.
• The proposed model showed significant improvements over existing YOLO detection models.

Conclusions:

• The Tea-YOLOv8s model offers enhanced detection precision for tea buds, despite a slight increase in parameters and computation.
• The model shows strong potential for application in automated tea harvesting equipment.
• Further research can explore optimizations for real-world deployment.